Clinical Involvement in Algorithm Development
Michael Pencina, vice dean of data science at the Duke University School of Medicine, says recent discussions about data-driven automation have focused on how to decrease burnout and increase efficiency by helping clinicians do what they want and spend less time worrying about administrative tasks.
He adds that it’s very important for clinicians to be involved in the planning of automation initiatives and the development of the algorithms themselves.
“The tool must be fit for purpose, address a real clinical operational need and be created with the practitioners, who are contributing throughout the process,” he says. “This isn’t a data science exercise; it must be created out of a need rather than just for the sake of technology.”
He adds that his team spends a lot of time identifying the top-priority use cases for digital innovation and AI technology.
“We are prioritizing the low-hanging fruit — where the risks are low and the gains can help in reducing burnout, improving the patient experience and, ultimately, improving the patient’s health,” says Pencina.
WATCH: Clinical automation offers relief amid staff shortages and rising burnout.
Analyzing Patient Behavior Using Predictive Modeling and Automation
Data-drive automation can also give insight into patient behavior, allowing healthcare organizations to make better operational decisions. Myers points to models that can analyze past patient behavior and determine whether a patient is going to be a no-show for an appointment.
“If you know a patient frequently does not turn up for their appointment, you can predict that and also double book the appointment,” she says. “This is where you can use predictive modeling and automation to ensure all of the patient slots are ultimately being filled, which ties back to revenue.”
For health systems seeking to implement data-driven clinical automation solutions, he recommends focusing on data governance, data transparency and identification of the areas that will benefit most.
“Governance means making sure it’s not just any algorithm, but it’s evaluated, and it’s done in a risk-based manner,” he explains. “Transparency is knowing what’s in place and understanding where the tools are coming from and what data they’ve been trained on.”
Pencina points out that it may not be beneficial — and would likely be impossible — to transform every area through automation all at once.
“Focus on identifying the use cases and measuring the progress,” he advises. “I think the stakes are high, but the opportunity is absolutely there.”
